I guess this epiphany meant the company did some progress on how to deploy and run their Kubernetes clusters, but the blog post reads like "we were messing our deployment strategy really badly and causing us a whole lot of problems for no good reason". I mean, calling their setup "chaos monkey" sounds like a desperate attempt to whitewash the problems they were creating for themselves.
And what was the magic ingredient? Switching to committed use nodes that can give customers up to a 70% discount? Was this supposed to be surprising?
I mean, take any intro course on the fundamentals of any cloud provider. Cost management is always one of the first topics covered. The difference between spot instance and committed pricing is rendered quite obvious.
The cloud 101 recommendation was always committed instances to cover baseline, and spin up any other instance type, perhaps spot instances, to handle peaks.
Is this what they are celebrating? That someone at the company finally paid attention to a cloud 101 course?
I guess this epiphany meant the company did some progress on how to deploy and run their Kubernetes clusters, but the blog post reads like "we were messing our deployment strategy really badly and causing us a whole lot of problems for no good reason". I mean, calling their setup "chaos monkey" sounds like a desperate attempt to whitewash the problems they were creating for themselves.
And what was the magic ingredient? Switching to committed use nodes that can give customers up to a 70% discount? Was this supposed to be surprising?
I mean, take any intro course on the fundamentals of any cloud provider. Cost management is always one of the first topics covered. The difference between spot instance and committed pricing is rendered quite obvious.
The cloud 101 recommendation was always committed instances to cover baseline, and spin up any other instance type, perhaps spot instances, to handle peaks.
Is this what they are celebrating? That someone at the company finally paid attention to a cloud 101 course?